The purpose of these experiments with octopuses (primarily ornatus) is to broaden our knowledge of invertebrate learning, which is now quite limited. Octopuses are chosen because they afford a marked evolutionary contrast with honeybees -- the invertebrates about whose learning most is known -- and because their sensory, motor, and motivational properties are suitable for work at the same high technical level. Locomotor and consummatory responses are classically and instrumentally conditioned by previously developed but hitherto unexploited techniques in a variety of experiments that already have shown a wide range of similarities between the learning of honeybees and vertebrates, similarities which are surprising in view of the remoteness of common ancestry and vast differences in brain size and organization. Particular attention is given to the paradoxical reward effects, which provide the first substantial evidence of divergence in vertebrate learning and of vertebrate-invertebrate convergence. Similarities and differences in the results for octopuses and other animals are expected to guide the search for underlying principles and mechanisms of learning, at least some of which may prove to be common to all animals, including humans.